Arboretum is a 134.6M sample dataset designed to advance AI for biodiversity applications by providing a large-scale, accurately annotated multimodal dataset that includes images and corresponding textual descriptions for a diverse set of species. Arboretum aims to facilitate the development of AI models for species identification, ecological monitoring, and agricultural research. Additionally, we introduce three new benchmark datasets: Arboretum-Unseen, Arboretum-LifeStages, and Arboretum-Balanced. As the authors of this submission, we affirm that we bear all responsibility in case of any rights violations or ethical issues associated with this work. We confirm that the submitted work is original, and if it includes third-party content, it is used with proper permissions and attributions.

We introduce BioTrove, the largest publicly accessible dataset designed to advance AI applications in biodiversity. Curated from the iNaturalist platform and vetted to include only research-grade data, BioTrove contains 161.9 million images, offering unprecedented scale and diversity from three primary kingdoms: Animalia ("animals"), Fungi ("fungi"), and Plantae ("plants"), spanning approximately 366.6K species. Each image is annotated with scientific names, taxonomic hierarchies, and common names, providing rich metadata to support accurate AI model development across diverse species and ecosystems. We demonstrate the value of BioTrove by releasing a suite of CLIP models trained using a subset of 40 million captioned images, known as BioTrove-Train. This subset focuses on seven categories within the dataset that are underrepresented in standard image recognition models, selected for their critical role in biodiversity and agriculture: Aves ("birds"), Arachnida ("spiders/ticks/mites"), Insecta ("insects"), Plantae ("plants"), Fungi ("fungi"), Mollusca ("snails"), and Reptilia ("snakes/lizards"). To support rigorous assessment, we introduce several new benchmarks and report model accuracy for zero-shot learning across life stages, rare species, confounding species, and multiple taxonomic levels. We anticipate that BioTrove will spur the development of AI models capable of supporting digital tools for pest control, crop monitoring, biodiversity assessment, and environmental conservation. These advancements are crucial for ensuring food security, preserving ecosystems, and mitigating the impacts of climate change. BioTrove is publicly available, easily accessible, and ready for immediate use.

Hence, we consider asking a VLM to provide the scientific name of the organism shown in a given image. There are two types of questions that we consider for this task. First, we consider open-ended questions, where we do not provide any answer choices (or options) to the VLM in the input prompt.

The top two rows show easy cases, while the bottom three rows present hard cases, including crowdedness, complex backgrounds, and tiny objects.

Fish-10K was manually curated with the help of biological experts co-authored in this paper.

Arboretum is a 134.6M sample dataset designed to advance AI for biodiversity applications by providing a large-scale, accurately annotated multimodal dataset that includes images and corresponding Arboretum aims to facilitate the development of AI models for species identification, ecological monitoring, and agricultural research. The dataset is hosted on Hugging Face. Our dataset will be available for as long as the iNaturalist Open Dataset is maintained.

Foundation models capable of generalizing across species and tasks represent a promising new frontier in bioacoustics, with NatureLM being one of the most prominent examples. While its domain-specific fine-tuning yields strong performance on bioacoustic benchmarks, we observe that it also introduces trade-offs in instruction-following flexibility. For instance, NatureLM achieves high accuracy when prompted for either the common or scientific name individually, but its accuracy drops significantly when both are requested in a single prompt. We address this by applying a simple model merging strategy that interpolates NatureLM with its base language model, recovering instruction-following capabilities with minimal loss of domain expertise. Finally, we show that the merged model exhibits markedly stronger zero-shot generalization, achieving over a 200% relative improvement and setting a new state-of-the-art in closed-set zero-shot classification of unseen species.

Fish-10K was manually curated with the help of biological experts co-authored in this paper.

Hence, we consider asking a VLM to provide the scientific name of the organism shown in a given image. There are two types of questions that we consider for this task. First, we consider open-ended questions, where we do not provide any answer choices (or options) to the VLM in the input prompt.